Skip to content

Commit

Permalink
Merge pull request #1 from shalb/hugging-face-model
Browse files Browse the repository at this point in the history
Hugging face model
  • Loading branch information
romanprog authored Nov 1, 2023
2 parents 618eff0 + 64d47bd commit 3e226d7
Show file tree
Hide file tree
Showing 15 changed files with 786 additions and 0 deletions.
30 changes: 30 additions & 0 deletions .github/workflows/release.yaml
Original file line number Diff line number Diff line change
@@ -0,0 +1,30 @@
name: Release

on:
push:
tags:
- '*'

env:
REGISTRY_URL: registry-1.docker.io
REGISTRY_REPO: shalb/charts

jobs:
release:
name: Release helm charts
runs-on: ubuntu-latest
container: alpine/helm:3.12.0
steps:
- uses: actions/checkout@v2

- name: Set env
run: |
echo "RELEASE_VERSION=${GITHUB_REF#refs/*/}" >> $GITHUB_ENV
- name: Build and push
run: |
echo ${{ secrets.DOCKERHUB_TOKEN }} | helm registry login -u ${{ secrets.DOCKERHUB_USERNAME }} --password-stdin ${{ env.REGISTRY_URL}}
for CHART_NAME in $(find -maxdepth 2 -type f -name "Chart.yaml" | cut -d"/" -f2); do
helm package --version ${{ env.RELEASE_VERSION }} ${CHART_NAME}
helm push ${CHART_NAME}-${{ env.RELEASE_VERSION }}.tgz oci://${{ env.REGISTRY_URL}}/${{ env.REGISTRY_REPO}}
done
23 changes: 23 additions & 0 deletions huggingface-model/.helmignore
Original file line number Diff line number Diff line change
@@ -0,0 +1,23 @@
# Patterns to ignore when building packages.
# This supports shell glob matching, relative path matching, and
# negation (prefixed with !). Only one pattern per line.
.DS_Store
# Common VCS dirs
.git/
.gitignore
.bzr/
.bzrignore
.hg/
.hgignore
.svn/
# Common backup files
*.swp
*.bak
*.tmp
*.orig
*~
# Various IDEs
.project
.idea/
*.tmproj
.vscode/
8 changes: 8 additions & 0 deletions huggingface-model/Chart.yaml
Original file line number Diff line number Diff line change
@@ -0,0 +1,8 @@
apiVersion: v2
name: huggingface-model
description: Helm chart for deploy Hugging Face to kubernetes cluster. See [Hugging Face models](https://huggingface.co/models)

type: application

version: 0.0.1

60 changes: 60 additions & 0 deletions huggingface-model/README.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,60 @@
# Helm chart for deploy Hugging Face to kubernetes cluster

See [Hugging Face models](https://huggingface.co/models)

## Parameters

### Model

| Name | Description | Value |
| --------------------------- | ---------------------------------------------------- | ----------------------------------------------------- |
| `model.organization` | Models' company name on huggingface, required! | `""` |
| `model.name` | Models' name on huggingface, required! | `""` |
| `init.s3.enabled` | Turn on/off s3 data source Default: disabled | `false` |
| `init.s3.bucketURL` | Full s3 URL included path to model's folder | `s3://k8s-model-zephyr/llm/deployment/segmind/SSD-1B` |
| `huggingface.containerPort` | Deployment/StatefulSet ContainerPort, optional | `8080` |
| `huggingface.args` | Additional arg for text-generation-launcher optional | `[]` |

### Global

| Name | Description | Value |
| --------------------------------- | ------------------------------------------------------------------------------------------------ | ----------------------------------------------- |
| `replicaCount` | Deployment/StatefulSet replicaCount | `1` |
| `kind` | Resource king [allowed values: deployment/StatefulSet, optional] | `deployment` |
| `image.repo` | Huggingface image repo | `ghcr.io/huggingface/text-generation-inference` |
| `image.tag` | Huggingface image version | `latest` |
| `image.pullPolicy` | Huggingface image pull policy | `IfNotPresent` |
| `imagePullSecrets` | May need if used private repo as a cache for image ghcr.io/huggingface/text-generation-inference | `[]` |
| `nameOverride` | String to partially override common.names.name | `""` |
| `fullnameOverride` | String to fully override common.names.fullname | `""` |
| `persistence.accessModes` | PVC accessModes | `["ReadWriteOnce"]` |
| `persistence.storageClassName` | Kubernetes storageClass name | `gp2` |
| `persistence.storage` | Volume size | `100Gi` |
| `service.port` | Service port, default 8080 | `8080` |
| `service.type` | Service type, default ClusterIP | `ClusterIP` |
| `serviceAccount.create` | Enable/disable service account, default enabled | `true` |
| `serviceAccount.role` | Kubernetes role configuration, default nil | `{}` |
| `podAnnotations` | Annotations for Redis® replicas pods | `{}` |
| `securityContext` | Set pod's Security Context fsGroup | `{}` |
| `extraEnvVars` | Array with extra environment variables to add to main pod | `[]` |
| `ingresses.enabled` | Enable/disable ingress(es) for model API, default disabled | `false` |
| `ingresses.configs` | List of ingresses configs | `[]` |
| `livenessProbe` | Configure extra options for model liveness probe | `{}` |
| `readinessProbe` | Configure extra options for model readiness probe | `{}` |
| `startupProbe` | Configure extra options for model startup probe | `{}` |
| `pdb.create` | Specifies whether a PodDisruptionBudget should be created | `false` |
| `pdb.minAvailable` | Min number of pods that must still be available after the eviction | `1` |
| `pdb.maxUnavailable` | Max number of pods that can be unavailable after the eviction | `""` |
| `resources.limits.nvidia.com/gpu` | The required option by text-generation-launcher | `1` |
| `resources.requests.cpu` | The requested CPU minimal recommended value | `3` |
| `resources.requests.memory` | The requested memory minimal recommended size | `10Gi` |
| `extraVolumes` | Optionally specify extra list of additional volumes for models' pods | `[]` |
| `extraVolumeMounts` | Optionally specify extra list of additional volumeMounts for models' container | `[]` |
| `autoscaling.enabled` | Enable Horizontal POD autoscaling for model | `true` |
| `autoscaling.minReplicas` | Minimum number of model replicas | `1` |
| `autoscaling.maxReplicas` | Maximum number of model replicas | `5` |
| `autoscaling.targetCPU` | Target CPU utilization percentage | `50` |
| `autoscaling.targetMemory` | Target Memory utilization percentage | `50` |
| `affinity` | Affinity for pod assignment | `{}` |
| `nodeSelector` | Node labels for pod assignment | `{}` |
| `tolerations` | Tolerations for pod assignment | `[]` |
83 changes: 83 additions & 0 deletions huggingface-model/templates/_helpers.tpl
Original file line number Diff line number Diff line change
@@ -0,0 +1,83 @@
{{/*
Generate internal container port.
*/}}
{{- define "huggingface-model.containerPort" -}}
{{- if .Values.huggingface }}
{{- default 8080 .Values.huggingface.containerPort }}
{{- else }}
8080
{{- end }}
{{- end}}
{{- define "common.capabilities.kubeVersion" -}}
{{- if .Values.global }}
{{- if .Values.global.kubeVersion }}
{{- .Values.global.kubeVersion -}}
{{- else }}
{{- default .Capabilities.KubeVersion.Version .Values.kubeVersion -}}
{{- end -}}
{{- else }}
{{- default .Capabilities.KubeVersion.Version .Values.kubeVersion -}}
{{- end -}}
{{- end -}}
{{/*
Expand the name of the chart.
*/}}
{{- define "huggingface-model.name" -}}
{{- default .Chart.Name .Values.nameOverride | trunc 63 | trimSuffix "-" }}
{{- end }}

{{/*
Create a default fully qualified app name.
We truncate at 63 chars because some Kubernetes name fields are limited to this (by the DNS naming spec).
If release name contains chart name it will be used as a full name.
*/}}
{{- define "huggingface-model.fullname" -}}
{{- if .Values.fullnameOverride }}
{{- .Values.fullnameOverride | trunc 63 | trimSuffix "-" }}
{{- else }}
{{- $name := default .Chart.Name .Values.nameOverride }}
{{- if contains $name .Release.Name }}
{{- .Release.Name | trunc 63 | trimSuffix "-" }}
{{- else }}
{{- printf "%s-%s" .Release.Name $name | trunc 63 | trimSuffix "-" }}
{{- end }}
{{- end }}
{{- end }}

{{/*
Create chart name and version as used by the chart label.
*/}}
{{- define "huggingface-model.chart" -}}
{{- printf "%s-%s" .Chart.Name .Chart.Version | replace "+" "_" | trunc 63 | trimSuffix "-" }}
{{- end }}

{{/*
Common labels
*/}}
{{- define "huggingface-model.labels" -}}
helm.sh/chart: {{ include "huggingface-model.chart" . }}
{{ include "huggingface-model.selectorLabels" . }}
{{- if .Chart.AppVersion }}
app.kubernetes.io/version: {{ .Chart.AppVersion | quote }}
{{- end }}
app.kubernetes.io/managed-by: {{ .Release.Service }}
{{- end }}

{{/*
Selector labels
*/}}
{{- define "huggingface-model.selectorLabels" -}}
app.kubernetes.io/name: {{ include "huggingface-model.name" . }}
app.kubernetes.io/instance: {{ .Release.Name }}
{{- end }}

{{/*
Create the name of the service account to use
*/}}
{{- define "huggingface-model.serviceAccountName" -}}
{{- if .Values.serviceAccount.create }}
{{- default (include "huggingface-model.fullname" .) .Values.serviceAccount.name }}
{{- else }}
{{- default "default" .Values.serviceAccount.name }}
{{- end }}
{{- end }}
125 changes: 125 additions & 0 deletions huggingface-model/templates/application.yaml
Original file line number Diff line number Diff line change
@@ -0,0 +1,125 @@
{{- if or ( not .Values.model ) ( or (not .Values.model.name) (not .Values.model.organization) ) -}}
{{- fail "model.name and model.id are required but not set!" -}}
{{- end }}
{{- $kind := .Values.kind | default "deployment" }}
{{- $namePrefix := include "huggingface-model.fullname" . | trunc 63 | trimSuffix "-" -}}
apiVersion: apps/v1
kind: {{ $kind }}
metadata:
annotations:
reloader.stakater.com/auto: "true"
name: {{ include "huggingface-model.fullname" . }}
labels:
{{- include "huggingface-model.labels" . | nindent 4 }}
spec:
replicas: {{ .Values.replicaCount }}
selector:
matchLabels:
{{- include "huggingface-model.selectorLabels" . | nindent 6 }}
template:
metadata:
{{- with .Values.podAnnotations }}
annotations:
{{- toYaml . | nindent 8 }}
{{- end }}
labels:
{{- include "huggingface-model.selectorLabels" . | nindent 8 }}
{{- if .Values.pdb.create }}
pdbEnabled: {{ include "huggingface-model.fullname" . }}
{{- end }}
spec:
{{- with .Values.imagePullSecrets }}
imagePullSecrets:
{{- toYaml . | nindent 8 }}
{{- end }}
{{- if eq $kind "StatefulSet" }}
serviceName: {{ include "huggingface-model.fullname" . }}
{{- end }}
{{- if .Values.serviceAccount.create }}
serviceAccountName: {{ include "huggingface-model.serviceAccountName" . }}
{{- end }}
{{- with .Values.podSecurityContext }}
securityContext:
{{- toYaml .Values.podSecurityContext | nindent 8 }}
{{- end }}
initContainers:
- name: init
image: alpine:3.18.4
imagePullPolicy: "IfNotPresent"
command: [ "/bin/sh", "-c" ]
args:
- if [ -d "/data/{{ .Values.model.name }}" ]; then echo "Model {{ .Values.model.id }} is already downloaded. Skipping init..."; exit 0; fi
{{- if and .Values.init.s3.enabled }}
- apk add --update aws-cli
- aws s3 cp --recursive {{ .Values.init.s3.bucketURL }} /data/{{ .Values.model.name }}
{{- else }}
- apk add --update git-lfs
- git clone --depth=1 https://huggingface.co/{{ .Values.model.id }}
{{- end }}
volumeMounts:
- name: model-storage
mountPath: /data
resources:
requests:
cpu: "3"
containers:
- name: model
image: {{ .Values.image.repo }}:{{ .Values.image.tag }}
imagePullPolicy: {{ .Values.image.pullPolicy | default "IfNotPresent" }}
command: [ "text-generation-launcher" ]
args:
- "--model-id"
- "{{ .Values.model.id }}"
- "--huggingface-hub-cache"
- "/usr/src/{{ .Values.model.name }}"
- "--weights-cache-override"
- "/usr/src/{{ .Values.model.name }}"
{{- with .Values.huggingface.args }}
{{ toYaml . | indent 12 }}
{{- end }}
ports:
- containerPort: {{ include "huggingface-model.containerPort" . }}
env:
- name: PORT
value: {{ include "huggingface-model.containerPort" . }}
- name: HUGGINGFACE_OFFLINE
value: "1"
{{- with .Values.extraEnvVars }}
{{ toYaml . | nindent 12 }}
{{- end}}
{{- with .Values.livenessProbe }}
livenessProbe: {{ toYaml . | nindent 12 }}
{{- end }}
{{- with .Values.readinessProbe }}
readinessProbe: {{ toYaml . | nindent 12 }}
{{- end }}
{{- with .Values.startupProbe }}
startupProbe: {{ toYaml . | nindent 12 }}
{{- end }}
volumeMounts:
- name: model-storage
mountPath: /data
{{- with .Values.extraVolumeMounts}}
{{- toYaml . | nindent 12 }}
{{- end }}
resources:
{{- toYaml .Values.resources | nindent 12 }}
{{- with .Values.nodeSelector }}
nodeSelector:
{{- toYaml . | nindent 8 }}
{{- end }}
{{- with .Values.affinity }}
affinity:
{{- toYaml . | nindent 8 }}
{{- end }}
{{- with .Values.tolerations }}
tolerations:
{{- toYaml . | nindent 8 }}
{{- end }}
volumes:
- name: model-storage
persistentVolumeClaim:
claimName: {{ include "huggingface-model.fullname" . }}
{{- with .Values.extraVolumes }}
{{- toYaml . | nindent 8 }}
{{- end }}
9 changes: 9 additions & 0 deletions huggingface-model/templates/auth-secret.yaml
Original file line number Diff line number Diff line change
@@ -0,0 +1,9 @@
{{- if .Values.authSecret }}
apiVersion: v1
data:
auth: {{ .Values.authSecret.basicAuth.authHash }}
kind: Secret
metadata:
name: {{ .Values.authSecret.basicAuth.secretName }}
{{- end }}

42 changes: 42 additions & 0 deletions huggingface-model/templates/hpa.yaml
Original file line number Diff line number Diff line change
@@ -0,0 +1,42 @@
{{- if .Values.autoscaling.enabled }}
{{- $kind := .Values.kind | default "deployment" }}
apiVersion: autoscaling/v2
kind: HorizontalPodAutoscaler
metadata:
name: {{ include "huggingface-model.fullname" . }}
labels:
{{- include "huggingface-model.labels" . | nindent 4 }}
spec:
scaleTargetRef:
apiVersion: apps/v1
kind: {{ $kind }}
name: {{ include "huggingface-model.fullname" . }}
minReplicas: {{ .Values.autoscaling.minReplicas }}
maxReplicas: {{ .Values.autoscaling.maxReplicas }}
metrics:
{{- if .Values.autoscaling.targetMemory }}
- type: Resource
resource:
name: memory
{{- if semverCompare "<1.23-0" (include "common.capabilities.kubeVersion" .) }}
targetAverageUtilization: {{ .Values.autoscaling.targetMemory }}
{{- else }}
target:
type: Utilization
averageUtilization: {{ .Values.autoscaling.targetMemory }}
{{- end }}
{{- end }}
{{- if .Values.autoscaling.targetCPU }}
- type: Resource
resource:
name: cpu
{{- if semverCompare "<1.23-0" (include "common.capabilities.kubeVersion" .) }}
targetAverageUtilization: {{ .Values.autoscaling.targetCPU }}
{{- else }}
target:
type: Utilization
averageUtilization: {{ .Values.autoscaling.targetCPU }}
{{- end }}
{{- end }}

{{- end }}
Loading

0 comments on commit 3e226d7

Please sign in to comment.